Xchool · Internal field note
The right way to adopt AI
Seat counts are theatre. True adoption is proof that named work moved.
Expose the top slice. Publish what they build. Put the rest of the work in the background. Only count proof the work moved.
The lie
Login means adopted.
Prompt class makes everyone fluent.
Power users will teach for free.
A bigger tool story is the product.
The fact
People do not want a tool that helps them get the work done.
They want the work done.
Using AI and using it well are different skills. Bad use makes the work worse than before.
The barbell is the design, not the failure
A perfect rollout still produces this shape. Build for it on day one. If a session needs everyone to become a craftsperson, cut it or split it.
Four moves. That is the whole method.
Train to expose, not to remediate
Every cohort answers: who is dormant, who can apply, who can multiply. Champions come from the instrument, not from nominations and hope.
Pay fluent people in status to publish
Their leverage is the gap. Close it only if teaching is public, ranked, and scoped. Private heroics keep a quiet aristocracy.
Finish the work in the tools they already use
Approve, edit, reject. Recurring packs, not agent school. Background completion beats mass prompt literacy.
Count work state and proof, never seats
For each named process: still manual, hybrid, or redesigned. Then show the artifact. Layer 1 without 2 and 3 is consulting cosplay.
What we measure
Skill
Who can do what. Spectrum, not a login flag.
Work
Share of named work still manual, hybrid, or moved.
Proof
Before and after. Learner output. Something a buyer can hold.
Never sell this
- AI adoption as the outcome. That metric is how boards get lied to.
- Prompt school as the product. Tips are table stakes. Judgment and proof are the work.
- Everyone becomes a builder. That future does not arrive. Do not price the offer on it.
If a buyer obsesses over adoption dashboards, reframe the score in discovery or accept you will be graded on theatre. Do not build delivery around a number you already know is fake.